Could your screening process spot a deepfake – or would it let it pass?

By Guest Writer | September 10, 2025 | Other

The rise of generative AI is reshaping how we work, communicate and how fraud is committed.

Employers are no longer just facing exaggerated or edited CVs, candidates are increasingly attending interviews through AI-generated avatars. The signs can be subtle: lip-sync errors, distorted audio, or a face that doesn’t quite look real.

These cases aren’t rare. Research shows that 85% of identity fraud attempts in 2024 involved AI-generated or manipulated content. Fraudsters are exploiting deepfakes, synthetic identities, and tampered documents to target recruitment processes, particularly where onboarding is remote or digital.

The risks may be evolving, but the consequences are familiar:

• Regulatory breaches if identity or Right to Work checks are compromised
• Safeguarding concerns if unverified individuals gain access to sensitive roles
• Operational disruption when fraud is uncovered late in the process
• Loss of trust, both within your organisation and with prospective employees

This isn’t just a concern for regulated industries, any employer hiring remotely could be exposed if they simply rely on visual checks and are not adopting technology.

Traditional screening wasn’t built for this

For years, identity verification has relied on whether a document looks genuine or if a face matches a photo. But when both can be convincingly manipulated with AI, those traditional methods are no longer enough.

The purpose of screening is simple: confirm that someone is who they say they are. Achieving that today means moving beyond appearance-based checks and building in technology that actively detects fraud, rather than relying on document uploads or video calls alone.

What effective screening looks like today

Modern screening must be able to keep pace with modern threats. That includes:

• Liveness detection to confirm a real person is present
• Biometric facial matching against trusted records
• Tamper detection on documents, including MRZ validation
• Deepfake identification capable of spotting facial morphing and synthetic injection
• Adaptable systems that evolve alongside new fraud techniques

When implemented well, these tools don’t slow down the process. In fact, they can speed it up, automatically approving genuine candidates while flagging only the small number of higher-risk cases for further review

It has to work for candidates too

As fraud becomes more advanced, so do candidate expectations. Screening must not only be secure, but also user-friendly, mobile-ready, and inclusive.

Over 50% of jobseekers say a poor onboarding experience has influenced their decision to reject a role. For many, identity verification is the very first meaningful interaction with an employer. If the process feels clunky or confusing, that impression lasts.

That’s why effective screening needs to act as a well-designed gateway, seamless across devices, supportive for non-native English speakers, and accessible for all candidates.

What comes next?

Fraud is not standing still, and neither can employers.

Protecting compliance is just the starting point. Organisations also need to safeguard their culture, reputation, and people. That protection begins with accurate identity verification at the very first step of onboarding.

So, it’s worth asking:

• Are your current checks designed to handle today’s risks?
• Can they reliably tell a genuine applicant from an AI-generated fake?
• Do you trust the very first decision you make – who you allow into your organisation?

If the answer isn’t a confident yes, it’s time to act.

This is a guest blog written by Mathew Armstrong, CEO of Giant Screening. Giant Screening is a UK-based background screening provider and trusted partner to some of the country’s best-known workplaces. Giant Screening is an accredited member of the Professional Background Screening Association (PBSA). Mathew contributes regularly on screening and onboarding standards, compliance, and the candidate experience.